forked from ROCm/pytorch
-
Notifications
You must be signed in to change notification settings - Fork 0
/
acos_op.cc
75 lines (64 loc) · 1.83 KB
/
acos_op.cc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
#include "caffe2/operators/acos_op.h"
#include "caffe2/utils/eigen_utils.h"
#include <algorithm>
#include <functional>
namespace caffe2 {
template <>
template <typename T>
bool AcosGradientFunctor<CPUContext>::Forward(
const std::vector<int>& X_dims,
const std::vector<int>& /* dY_dims */,
const T* X,
const T* dY,
T* dX,
CPUContext* /* context */) const {
const int size = std::accumulate(
// NOLINTNEXTLINE(modernize-use-transparent-functors)
X_dims.cbegin(), X_dims.cend(), 1, std::multiplies<int>());
ConstEigenVectorArrayMap<T> dY_arr(dY, size);
ConstEigenVectorArrayMap<T> X_arr(X, size);
EigenVectorMap<T>(dX, size) = -dY_arr * (T(1) - X_arr.square()).rsqrt();
return true;
}
REGISTER_CPU_OPERATOR(
Acos,
UnaryElementwiseOp<
TensorTypes<float>,
CPUContext,
AcosFunctor<CPUContext>>);
REGISTER_CPU_OPERATOR(
AcosGradient,
BinaryElementwiseOp<
TensorTypes<float>,
CPUContext,
AcosGradientFunctor<CPUContext>>);
OPERATOR_SCHEMA(Acos)
.NumInputs(1)
.NumOutputs(1)
.IdenticalTypeAndShape()
.SetDoc(R"DOC(
Calculates the arccosine of the given input tensor, element-wise.
)DOC")
.Input(0, "input", "Input tensor")
.Output(
0,
"output",
"The arccosine of the input tensor computed element-wise");
OPERATOR_SCHEMA(AcosGradient)
.NumInputs(2)
.NumOutputs(1)
.IdenticalTypeAndShape();
namespace {
class GetAcosGradient : public GradientMakerBase {
using GradientMakerBase::GradientMakerBase;
std::vector<OperatorDef> GetGradientDefs() override {
return SingleGradientDef(
"AcosGradient",
"",
std::vector<std::string>{I(0), GO(0)},
std::vector<std::string>{GI(0)});
}
};
} // namespace
REGISTER_GRADIENT(Acos, GetAcosGradient);
} // namespace caffe2